基于补充频繁模式的P2P搜索优化

为有效提高peer to peer(P2P)系统在文件共享应用中的资源搜索速度，解决P2P节点的负载均衡，保证P2P共享系统的稳定性、可靠性和扩展性，提出了一种基于补充频繁模式的快速查询响应方法，主要工作包括：（1）提出满事务和补充频繁模式概念及补充频繁模式树(SFP-Tree)结构，其主要思路是利用具有较高支持度计数的短频繁模式B，吸收与B一起构不成频繁模式但又达到一定共享支持度阀值的单项构成基项集的补充项，B的所有补充单项构成其补充项集S，在此基础上查找包含B的最大频繁项集X，并从S中排除与X一起达不到共享支持度阀值的补充项，从而以X和S的子集构成一个以B为基项集的具有足够长度的补充频繁模式； （2）实现了补充频繁模式的挖掘方法，其主要思路是在构成整个事务数据库升序FP-Tree(AFP-Tree)基础上，递归挖掘条件混合模式树（CMP-Tree），在基项达到指定长度时，生成补充频繁模式树SFP-Tree，在此基础上挖掘带有负项的补充频繁模式，最后针对带有负项的补充频繁模式生成只包含正项的补充频繁模式； （3）模拟一份文件资源的历史响应事务集进行详实的实验，实验结果表明对于数据集规模达到100，000的P2P系统历史响应事务数据库，基于SFP-Tree的补充频繁模式挖掘方法能在20秒以内生成此文件资源的有效响应列表并能选择其中一个列表的有效子集瞬时响应查询请求。

Search optimization of P2P system based on SFP-tree

To realize the rapidly search and load balance of file sharing in Peer to Peer system, this paper proposes the supplemental frequent pattern based sites distribution method to response share queries. The main contents include: (1) proposing the conception of full transaction, supplemental frequent pattern along with the structure of SFP-Tree. To a item set B with high support count, each item not being frequent with B but relative frequent (arrive to the share support threshold) becomes the supplemental item of B, and a frequent item set X including B can attract some supplemental items of B to form an item set with enough items. This type of item set is long enough and efficient enough for providing the partitions of relative file synchronously. (2) giving the method of supplemental frequent pattern mining using AFP-Tree, CMP-Tree and SFP-Tree, and (3) executing simulation experiment to test the efficiency of SFP-Tree based supplemental frequent pattern mining, the result shows that SFP-Tree bases method can realize the supplemental frequent pattern mining in 20 seconds to a query response transaction with 100,000 transactions.